Assessing risk perception by means of ordinal models 79The study focused on the blue-collar population of six different departments,each
dealing with a specific industrial process. The subjects in the cohort are distributed
among the following six units, whose main activities are also described. In the Plates
department, workers must set plates and cylinders used during the printing opera-
tions and then carried out in the Rotogravure and Offset departments. The Packaging
department is responsible for the bookbinding and packaging operations, while the
Plants department operates several systems (e.g., electrical and hydraulic) and pro-
vides services (e.g., storage and waste disposal) that support the production side of the
company. Lastly, the Maintenance department workers perform a series of operations
connected with the monitoring and correct functioning of the different equipment of
the plant.
With the purpose of studying injury risk perception among company workers,
a structured “Workplace Risk Perception Questionnaire” was developed. The ques-
tionnaire asked the respondents to express their opinions on a series of risk factors
present in their workplace. A 7-point Likert scale was used to elicit the workers’ an-
swers whose ranges are interpreted as: 1 = “low perceived risk”; 7 = “high perceived
risk”. Moreover, we pay particular attention to socio-demographic characteristics
like ‘gender’ (dichotomous variable ’0’=men and ’1’=women) , ‘number of working
years’ within the company (continuous variable ranging form 1 to 30) and ‘type of
injury’ (dichotomous variable ’0’= not severe injury and ’1’= severe injury). Finally,
n=348 validated questionnaires were collected.
3.2 Control and measure risk perception: a map
As already discussed, we built a class of model to evaluate, control and measure the
risk perception and, means of monitoring activity, to inform the stakeholders of the
direction of new policies. In this case we show a map of synthesis which contains
whole information related to different risk dimensions.
In Figure 1 we plot for each item the reactions of feeling and uncertainty expressed
by people. We can observe that the uncertainty is concentrated between 0 and 0.6, a
range indicating a high level of indecision. The characteristic of feeling, however, is
extended over the whole parametric space. Both aspectsillustrate how the responses
interact to determine behaviour. Moreover, we deepen some specific aspects of risk-
related phenomena that are regarded as more interesting.
In the case ofcontrol, for example, we can observe a dichotomous behaviour:
less sensitivity for injuries such as eye-wound, hit, moving machinery clash (in these
cases people do not seem to ask for more control), and more for other injuries where
people, on their scale of risk, consider the aspect of control as a sensible variable for
improving the conditions of their job.
Instead, an interesting evaluation is referred to astraining, as it is considered an
important variable of the survey. Less evidence appears for other items shared among
different levels and whose estimates are spread over the parametric space.